Isolation and identification of new inner membrane-associated proteins that localize to cell poles in Escherichia coli


E-mail; Tel. (+1) 501 526 6802; Fax (+1) 501 686 5359.


Several bacterial structures, processes and proteins are localized primarily to the poles of rod-shaped cells. To better understand this cellular organization, we devised a new method for identifying proteins that localize to the poles of Escherichia coli. Pole-derived membrane fragments were isolated by affinity capture of vesicles containing the chemotaxis protein, Tar; and for comparison, vesicles representing all parts of the cytoplasmic membrane were captured by expressing a Tar variant that was no longer pole-specific. A combination of one-dimensional SDS-PAGE and semi-quantitative mass spectrometry identified 31 proteins that were highly enriched in polar vesicles. Five were chemotaxis proteins known to be pole-specific and another, Aer, was an aerotaxis protein that had not yet been localized to the pole. The behaviour of these internal controls validated the overall approach. GFP-fused derivatives of four candidates (Aer, YqjD, TnaA and GroES) formed polar foci that were distinct from inclusion bodies. TnaA–GFP and GroES–GFP were functional, formed a single focus per cell, and competed for polar localization with the wild-type versions of these proteins. Polar localization of TnaA, GroES and YqjD was disrupted in cells lacking the MinCDE proteins, suggesting that this system may help localize proteins not involved in cell division.


All bacteria, even those that are spherical, display geometric asymmetries during at least part of their life cycle, and it appears that cells take advantage of these asymmetries to localize cellular components to specific sites. This tendency for selective localization is especially apparent at the poles of rod-shaped bacteria. For example, some membrane microdomains are enriched for individual lipids or lipid combinations, and these differences are especially conspicuous at the poles (Vanounou et al., 2003; Huang et al., 2006; Renner and Weibel, 2011). Escherichia coli also displays an intriguing micro-heterogeneity in the distribution of its membrane proteins (Jacoby and Young, 1988; 1990), and once again this effect is most clearly associated with the poles (Lai et al., 2004). Finally, peptidoglycan at the poles is, in some way, different from that in the cylindrical portions of the cell, in that polar peptidoglycan is naturally inert – that is, it is not degraded, recycled or diluted by addition of new material (Koch and Woldringh, 1994; de Pedro et al., 2003). This inert, or static, nature of the poles also encompasses several polar-localized proteins and lipopolysaccharides (de Pedro et al., 2004; Nilsen et al., 2004; Rudner and Losick, 2010).

In the past few years, a noteworthy number of proteins and macromolecular complexes have been observed to localize to bacterial poles. These include chemotaxis proteins, parts of the chromosomal segregation apparatus, type IV pili, slime nozzles in gliding bacteria, two-component signalling proteins, septation regulators, flagella, actin polymerization nucleators, autotransporters, type II secretion proteins, type III secretion systems, a type IV apparatus, an RNA helicase and a bacterial reverse transcriptase (Maddock and Shapiro, 1993; Gestwicki et al., 2000; Lai et al., 2000; Sourjik and Berg, 2000; Errington et al., 2001; 2003; Lybarger and Maddock, 2001; Scott et al., 2001; Kumar and Das, 2002; Shapiro et al., 2002; El-Fahmawi and Owttrim, 2003; Grohmann et al., 2003; Vignon et al., 2003; Wu and Errington, 2003; Janakiraman and Goldberg, 2004; Chakravortty et al., 2005; Judd et al., 2005; Zhao and Lambowitz, 2005). The breadth of the types of protein complexes exhibiting a polar association is impressive. Other examples of individual proteins that apparently prefer to localize to the poles include the E. coli LacY and ProP proteins (Romantsov et al., 2010), as do proteins of the phosphotransferase system (Lopian et al., 2010). In Bacillus subtilis, the SPP1 phage infects preferentially at the poles (Jakutyte et al., 2011), as does phage lambda in E. coli (Rothenberg et al., 2011). Competence and recombination proteins localize to the poles of B. subtilis, making this the preferred site for DNA binding and uptake (Hahn et al., 2005; Kidane and Graumann, 2005; Kaufenstein et al., 2011). Similar DNA transfer machinery is confined to the poles in Streptomyces sp. (Grohmann et al., 2003) and in Agrobacterium tumefaciens (Lai et al., 2000; Kumar and Das, 2002); and in Caulobacter crescentus a large number of polar proteins co-ordinate DNA segregation and cell cycle progression (Ebersbach et al., 2008; Schofield et al., 2010).

So far, the discovery of pole-localized proteins has been mostly a by-product of investigations into specific biological pathways and processes; few attempts have been made to build a systematic catalogue of proteins that localize specifically or preferentially to these sites. One exception is the work of Lai et al., who began to create a list of pole-specific components by identifying proteins that accumulated preferentially in E. coli minicells (Lai et al., 2004). Because each minicell is composed predominantly of polar material (half old and half new), Lai et al. predicted that proteins having a strong affinity for the poles would be enriched in these particles. Lai et al. screened 173 proteins from two-dimensional electrophoresis gels and found that 36 were enriched in minicells, although only 13 were deemed to be pole-specific candidates (Lai et al., 2004). Immunomicroscopy confirmed that one candidate (OmpW) was pole-specific, while another (YiaF) was distributed throughout the cell, albeit with some predilection for the poles (Lai et al., 2004). No follow-up investigations of other candidates have been reported.

Alternative approaches for screening for pole-specific proteins have involved expressing libraries of proteins fused to fluorescent domains (e.g. GFP or mCherry) that can be tracked by microscopy (Kitagawa et al., 2005; Janakiraman et al., 2009; Werner et al., 2009). For example, Kitagawa et al. constructed the ASKA library, in which E. coli open reading frames were cloned into plasmids so that each expressed protein displayed a poly-histidine tag at its N-terminus and a GFP tag at its C-terminus (Kitagawa et al., 2005), and Janakiraman et al. screened these for polar localization (Janakiraman et al., 2009). The fluorescent images of most of these fusion proteins are available online ( Werner et al. employed a similar strategy to identify polar proteins in C. crescentus, resulting in the identification of nearly 300 localized proteins (Werner et al., 2009). Bowman et al. used a bioinformatic screen for the same purpose (Bowman et al., 2008). These latter authors began with 46 known polar proteins, each involved in pilus assembly, flagellar assembly, polar signalling or which were components of polar organelles such as the cell stalk. Using these known polar proteins as bioinformatic ‘bait’, Bowman et al. predicted that 36 conserved hypothetical proteins interacted with them and were candidates for being new polar proteins. Screening for truly polar proteins yielded one, PopZ, that helps anchor the chromosomal origin to the poles of C. crescentus (Bowman et al., 2008).

However useful for identifying polar candidates, each of the above approaches has its own limitations. The strategy of Lai et al. was skewed strongly towards finding only polar-localized integral outer membrane proteins. Plasmid-encoded fusion proteins in the ASKA and Caulobacter libraries were expressed in E. coli and C. crescentus, respectively (Kitagawa et al., 2005; Werner et al., 2009), so that the tagged proteins competed with chromosomally encoded wild-type proteins for polar target sites. Such competition can prevent such tagged proteins from localizing to the pole, as was observed in the case of IcsA (Charles et al., 2001). Finally, the bioinformatics approach relies on known and predictable protein–protein interactions and may thus eliminate from consideration numerous unknown candidates (Bowman et al., 2008).

In this work, we report the results of a complementary strategy to isolate and identify proteins that localize to cell poles in E. coli. We fused the FLAG peptide tag to a known polar protein, Tar, and employed affinity chromatography to isolate inner membrane vesicles that originated primarily from the poles. For comparison, a truncated non-polar version of Tar-FLAG was used to isolate a general population of cytoplasmic membrane vesicles. Proteins associated with these vesicles were separated by gel chromatography and analysed by semi-quantitative mass spectrometry. We identified 31 candidate proteins that localized exclusively to polar vesicles or that were highly enriched in these vesicles. Four known polar proteins and four new candidates were tested and confirmed to localize to the poles in vivo, thus validating the overall approach. In addition, we found that the Min system plays a role in the polar localization of three of these proteins. Extension of this affinity-capture technique should help create a systematic list of pole-localized proteins after additional candidates are evaluated and by using other known polar proteins as affinity markers.


Isolation of pole-derived inner membrane vesicles

Our first objective was to devise a procedure to isolate unknown proteins that localized to the cell poles of E. coli. We reasoned that the cytoplasmic membrane would contain several such proteins and that these could be enriched by using a known polar protein to capture membrane vesicles derived specifically from this region. We chose the Tar protein as a polar marker because it is a relatively abundant inner membrane chemoreceptor that clusters with high specificity at the poles (Maddock and Shapiro, 1993; Shiomi et al., 2005; Greenfield et al., 2009). Also, the crystal structure of Tar facilitated the insertion of ectopic affinity tags without compromising protein function (Bowie et al., 1995; Kim et al., 1999). Tar spans the inner membrane twice, with one short N-terminal peptide in the cytoplasm, one looped periplasmic domain, and a C-terminal cytoplasmic domain. To create a polar vesicle tag, we engineered a plasmid to express Tar with a FLAG epitope inserted at the distal end of the periplasmic loop (Tar-P-FLAG) (Fig. 1A). This insertion did not disrupt the polar localization of Tar-P-FLAG, as determined by anti-FLAG immunolabelling (Fig. 1B). Because the FLAG tag was located in the periplasm, affinity capture of this Tar protein would allow us to isolate right-side-out membrane vesicles. We also fused FLAG to the cytoplasmic C-terminus of Tar (Tar-C-FLAG) (Fig. 1A). Because inner membrane vesicles prepared by mechanical disruption are often inside-out (Futai, 1974), this construct would allow us to isolate inverted vesicles. A third Tar derivative was constructed by deleting most of the cytoplasmic domain of Tar-P-FLAG and replacing it with GFP (Tar-P-FLAGΔ) (Fig. 1A). Removing the cytoplasmic sequence creates a membrane-bound Tar protein that no longer localizes specifically to the poles (Shiomi et al., 2006). As expected, Tar-P-FLAGΔ was distributed evenly around the circumference of the cells (Fig. 1B). This truncated Tar protein allowed us to isolate a general population of inner membrane vesicles, and the proteins in these vesicles provided a baseline for comparing pole-specific protein candidates.

Figure 1.

Construction and localization of Tar-FLAG proteins. A. Schematics of plasmid-encoded Tar proteins with FLAG insertion sites indicated (P, periplasmic; Cyto, cytoplasmic). Trans-inner membrane domains are shaded in light grey and the GFP domain is shaded in dark grey. Tar-P-FLAGΔ is a truncated protein in which the majority of the cytoplasmic domain has been replaced by GFP. B. Localization of Tar-FLAG proteins. Cells containing pTar-P-FLAG or pTar-P-FLAGΔ were grown in 1% Tryptone medium supplemented with 25 µM IPTG. The Tar-P-FLAG and vector control cells were immunolabelled with anti-FLAG antibodies conjugated with fluorescein isothiocyanate (FITC). Cells containing Tar-P-FLAGΔ were observed directly for GFP fluorescence. Upper panels, phase-contrast images; lower panels, fluorescence images.

To enrich pole-localized proteins by differential capture of membrane vesicles, we expressed the three Tar proteins separately in E. coli MG1655, disrupted the cells to enrich for small membrane vesicles, and captured subsets of these vesicles by affinity chromatography. Tar-FLAG tagged vesicles were isolated by incubating the lysate with anti-FLAG antibody attached to agarose beads, after which the proteins associated with or embedded within these vesicles were solubilized and released by incubation with Triton X-100. These proteins constituted a population enriched in pole-associated membrane vesicles. Any proteins remaining on the lipid-free beads were solubilized with hot SDS (mainly Tar-FLAG that was bound by anti-FLAG antibody). All samples were separated through one-dimensional SDS-PAGE and stained with Sypro Ruby (Fig. 2). The Triton X-100-eluted samples contained a wide range of proteins, while the SDS-eluted samples contained proteins corresponding to the Tar-FLAG proteins (Fig. 2, arrows). Equivalent protein profiles were observed when inside-out vesicles were captured by using Tar-C-FLAG (Fig. 2, lanes 5 and 9) or when right-side-out samples were captured by using Tar-P-FLAG (Fig. 2, lanes 4 and 8). These results showed that either Tar construct could be used to capture similar subpopulations of membrane vesicles, indicating that both tagged proteins could be used to isolate membranes derived from a common location. On the other hand, membrane vesicles tagged with Tar-P-FLAGΔ (non-polar) contained a different array of proteins (Fig. 2, lanes 3 and 7). Most notably, when compared with this control, several proteins were either enriched or absent in pole-derived vesicles captured by Tar-P-FLAG (Fig. 2, bands A, B, C and D). Thus, the protein content of putative polar vesicles was different from that of vesicles isolated from the overall inner membrane, suggesting that the procedure could capture vesicles derived from the pole.

Figure 2.

Protein profiles in polar and non-polar vesicles. Pole-derived vesicles labelled with Tar-P-FLAG or Tar-C-FLAG and non-pole-derived vesicles labelled with Tar-P-FLAGΔ were isolated by affinity capture by using anti-FLAG beads. Triton X-100- and SDS-eluted samples were separated on a 12% SDS-PAGE gel and stained with Sypro Ruby. Total inner membrane (IM) proteins from E. coli MG1655 cell envelopes were solubilized with Triton X-100 and loaded for comparison (lane 1). Protein bands to the left of each capital letter were removed and identified by mass spectrometry. (A) Tar, Tsr, Tap and Trg; (B) MukB; (C and D) Tar multimers. Tar-FLAG bands are indicated by arrows. M, gel lanes subjected to full-lane GeLC-MS/MS analysis.

Identification of proteins enriched in pole-derived membrane vesicles

Before identifying the proteins in each sample by mass spectrometry (MS), we wanted to confirm that the vesicles captured by using Tar-P-FLAG really represented a pole-derived population. To test this, we identified the proteins in three sets of bands that were clearly enriched in samples containing pole-derived (Fig. 2, lanes 4 and 8) or non-pole-derived (Fig. 2, lane 7) vesicles. These proteins were: (i) several proteins of about 60 kDa (A-band) that were enriched in polar vesicles and released by Triton X-100 (Fig. 2, lane 4), (ii) a band of ∼ 150 kDa (B-band) present in the SDS-eluted fraction of the non-polar sample (Fig. 2, lane 7), and (iii) proteins of ∼ 150 kDa (C-band) and ∼ 120 kDa (D-band) that were enriched in the SDS-eluted fraction of the polar sample (Fig. 2, lane 8). The A-band contained the four chemoreceptors Tar, Tsr, Tap and Trg. The presence of Tar in this fraction was expected because this protein was used as the polar marker. Simultaneous enrichment of the three related chemoreceptor proteins in these vesicles verified that the overall strategy was successful. By capturing vesicles containing Tar, the procedure also captured proteins closely associated with Tar and, by extension, any other proteins located in or associated with vesicles derived from the polar region. Both the C-band and D-band and were identified as being Tar, presumably dimers and multimers that were bound to anti-Tar antibody and released from the affinity beads only by SDS solubilization. The B-band contained the MukB protein, which segregated specifically into non-polar membrane vesicles in each of three trials (e.g. Fig. 2, lane 7, and data not shown). The behaviour of MukB is consistent with the fact that this protein localizes specifically to midcell in E. coli to help partition chromosomes to the two daughter cells (Danilova et al., 2007). Thus, MukB is normally located far from the poles of E. coli and should therefore be absent from polar membrane vesicles but present in vesicles representing the inner membrane at large. Because this was the case, MukB provided a good non-pole marker. In short, membrane vesicles that were captured because they contained a known polar protein, Tar, also contained related pole-localized chemoreceptors but did not contain the non-pole protein, MukB. On the other hand, vesicles representing the general cytoplasmic membrane contained the non-pole protein, MukB, but contained few or none of the pole-localized chemoreceptors. Together, these results indicated that the affinity capture procedure could, with a high degree of specificity, distinguish between polar and non-polar vesicles and that proteins colocalizing to the pole could be isolated in this way.

Having corroborated the effectiveness of the overall approach, we then identified other proteins associated with polar and non-polar membrane vesicles. Individual lanes from a replicate SDS-PAGE gel (as in Fig. 2) were subjected to full-lane GeLC-MS/MS. We confirmed that the Tar-FLAG proteins were present in affinity-purified membrane vesicles but that they were not present on affinity beads exposed to membrane vesicles from cells of the vector-only control. However, the vector control samples contained many proteins that bound tightly and non-specifically to the affinity beads and were eluted only by SDS treatment (Fig. 2, lane 6), so we did not analyse these proteins further. In contrast, when we eluted vesicle-associated proteins by solubilizing attached membranes with Triton X-100, the numbers of proteins captured in vesicles containing Tar-P-FLAGΔ or Tar-P-FLAG (Fig. 2, lanes 3 and 4) were much greater than the number of proteins present in vesicle fractions containing neither derivative (Fig. 2, lane 2). This confirmed that the former proteins were associated specifically with Tar-FLAG-marked vesicles, and these samples were investigated further.

The combination of one-dimensional SDS-PAGE plus GeLC-MS/MS identified a total of 124 proteins in polar and non-polar vesicles (Table S1 contains the complete list). These were identified by amassing 1940 spectra with a minimum peptide probability of 90%, a protein probability of 99%, and at least three spectra for each protein, not including the three most abundant contaminating proteins (AceF, AceE and LpdA) in the vector-only controls. Of these 124 proteins, 62 were integral inner membrane proteins and 50 others were known to be associated with the inner membrane or that have been detected in inner membrane fractions. Of particular importance was that of the proteins that co-purified with these vesicles only 12 were not known to associate with the inner membrane, indicating that the vesicle isolation procedure captured, almost exclusively, fragments of the inner membrane.

Proteins enriched in pole-localized vesicles were identified by a semi-quantitative approach in which the amount of protein in polar samples was divided by the amount of protein in the non-polar fractions, producing a ‘polar score’. In each sample, the relative quantity of a protein was represented by the number of spectra (spectrum count) assigned to individual proteins (Zybailov et al., 2005), and non-specific background counts on the affinity column were accounted for by subtracting the spectrum count of each protein in the vector control from those counts in polar and non-polar samples. The polar score was > 1 if the proteins were enriched in pole-derived vesicles, < 1 if the proteins were located more often in the sidewalls and not at the poles, and equal to 1 if the proteins were evenly distributed throughout both polar and sidewall membranes. The most abundant proteins identified in both samples were the F0F1 ATP synthase components AtpA, D, F, G and H. The ATP synthase proteins had combined net spectrum counts of 122 and 112 in the polar and non-polar samples, respectively (Table 1), consistent with the finding that AtpA–GFP and AtpD–GFP are distributed evenly throughout the cytoplasmic membrane and should therefore be present in equal amounts in polar and non-polar vesicles (Kitagawa et al., 2005). In contrast, the chemotaxis proteins (Tar, Tsr, Tap, Trg and CheA) had a combined net spectrum count of 254 in polar vesicles but only 26 in non-polar vesicles (Table 1), indicating that the semi-quantitative approach was able to detect and differentiate proteins enriched in one or the other sample.

Table 1.  Mass spectrometry results for proteins identified in polar and non-polar vesicles.
  • a. 

    Only Triton X-100-eluted proteins are included. See Table S1 for a complete list of proteins.

  • b. 

    Proteins identified only in polar vesicles.

  • c. 

    Combined net spectrum counts (NSC) of ATP synthase components (AtpA, D, F, G and H).

  • d. 

    Combined NSC of chemotaxis proteins (Tar, Tsr, Tap, Trg and CheA).

  • IM, inner membrane.

Number of proteinsa102124
Number of IM-associated proteins95112
Number of pole-specific proteinsb22
NSC of ATP synthasec112122
NSC of chemotaxis proteinsd26249

Twenty-two proteins were present only in polar vesicles but not in non-polar vesicles, and nine other proteins had a polar score of 5 or greater (Table 2). Overall, there were 26 pole-enriched proteins in addition to the five known chemotaxis proteins (Table 2). These 26 included the proteins PBP1B and Pal that are known to be present during septation and which possibly remain at the new poles after cell division (Bertsche et al., 2006; Gerding et al., 2007). The remaining 24 proteins had not been localized to poles in previous screens.

Table 2.  Proteins enriched in polar vesicles.
ProteinFunction and subcellular localizationaPolar scorebGFP-fusion phenotypec
  • a. 

    Subcellular locations are annotated according to Diaz-Mejia et al. (2009). C, cytoplasmic; IM, inner membrane; P, periplasmic; OM, outer membrane; EC, extracellular; IMA, IM-associated.

  • b. 

    Polar scores are the net spectrum counts of proteins identified only in polar vesicles, or the polar: non-polar ratio of the net spectrum counts of proteins identified in both polar and non-polar vesicles.

  • c. 

    ‘(−)’ indicates that no GFP fusion was constructed for this protein.

Identified only in polar vesicles
AerAerotaxis receptor, IM18Polar foci
PBP1BPenicillin-binding protein 1B, IM8(−) Known septal protein (Bertsche et al., 2006)
TnaATryptophanase, C IMA8Polar foci
DcuAAnaerobic C4-dicarboxylate transporter, IM7No signal
PutPSodium/proline symporter, IM6No signal, growth defect
TrxAThioredoxin-1, C IMA6(−)
CheAChemotaxis protein, C IMA5(−) Known polar protein (Sourjik and Berg, 2000)
FliPFlagellar biosynthetic protein, IM5No signal, growth defect
AccDAcetyl-CoA carboxyltransferase β, C IMA4(−)
CpxACpxAR signal transduction sensor protein, IM4No signal
FliCFlagellin, EC4No signal, growth defect
PalPeptidoglycan-associated lipoprotein, OM4(−) Known septal protein (Gerding et al., 2007)
RpsD30S ribosomal protein S4, C4(−)
YcbCUncharacterized protein, IM4No signal, growth defect
BamAOuter membrane protein assembly factor, OM3(−)
GlnPGlutamine transport system permease protein, IM3(−)
GroEL60 kDa chaperonin 1, C3Polar foci (GroES–GFP)
MlaFPhospholipid import ATP-binding protein, C IMA3No signal
NarZRespiratory nitrate reductase 2 alpha chain, P IMA3No signal, growth defect
TolCOuter membrane protein, OM IMA3(−)
YqjDUncharacterized protein, C IMA3Polar foci
YniBUncharacterized protein, IM3(−)
Identified in both polar and non-polar vesicles
TapMethyl-accepting chemotaxis protein IV, IM20(−) Known polar protein (Hazelbauer et al., 2008)
TrgMethyl-accepting chemotaxis protein III, IM16(−) Known polar protein (Hazelbauer et al., 2008)
YijPPhosphoethanolamine transferase, IM12No signal
PBP1APenicillin-binding protein 1A, IM9No signal
TsrMethyl-accepting chemotaxis protein I, IM9(−) Known polar protein (Ping et al., 2008)
TarMethyl-accepting chemotaxis protein, II7(−) Known polar protein (Greenfield et al., 2009)
SppAProtease 4, IM6.5(−)
LepBSignal peptidase I, IM6No signal
NupCNucleoside permease, IM5(−)

Aer, YqjD, TnaA and GroES GFP fusion proteins form polar foci

We wished to confirm that some of these proteins enriched in polar membranes were in fact localized to the poles. We chose 15 candidates that were highly enriched in polar vesicles but that had never been reported to be localized at cell poles (Table 2) and constructed C-terminal GFP fusions to the chromosomal copies of these genes, so that expression of each composite was controlled by its native promoter. An exception was the polar candidate GroEL. The crystal structure of GroEL indicates that its C-terminus is not exposed on the surface of the GroES–GroEL complex (Xu et al., 1997), meaning that it was not possible to fuse GFP to this site without destroying protein function. Therefore, as an alternative, we fused GFP to the surface-exposed C-terminus of GroES, which forms a hetero-multimeric complex with GroEL (Xu et al., 1997). For use as a non-polar control, we selected the cytoplasmic protein GapA, which was enriched only slightly in polar vesicles and therefore was expected to be distributed relatively evenly throughout the cytoplasm.

Four of fusion proteins (Aer, YqjD, TnaA and GroES) exhibited detectable GFP fluorescence by microscopy. The remaining 11 fusions showed only very weak fluorescence signals, and expression of seven of these produced severe growth defects (not shown). These 11 were therefore not pursued here. Aer–GFP, YqjD–GFP, TnaA–GFP and GroES–GFP produced distinct fluorescent foci that localized almost exclusively to the poles (Fig. 3A), whereas the GapA–GFP non-polar control was spread throughout the cytoplasm (Fig. 3B). Two polar patterns were observed. Aer–GFP and YqjD–GFP formed clusters of various sizes at different locations in the cells, with the largest clusters localized predominately at the poles. This distribution is typical for chemoreceptors (Greenfield et al., 2009), and is consistent with the fact that Aer is an aeroreceptor that interacts with chemoreceptors (Gosink et al., 2006). Therefore, Aer serves as another proof-of-concept protein, showing that the affinity procedure can identify pole-associated proteins. In contrast, TnaA–GFP and GroES–GFP formed only one fluorescent focus per cell, located either at one of the poles or at midcell (Fig. 3A), suggesting that the localization of these proteins differed from that of Aer and YqjD and might be more pole-specific.

Figure 3.

Aer, YqjD, TnaA and GroES GFP fusion proteins form polar foci in E. coli. A. Foci are not misfolded protein inclusion bodies. E. coli cells coexpressing one GFP fusion protein and IbpA–mCherry (to mark inclusion bodies) were grown at 37°C in 1% Tryptone medium plus kanamycin until the culture reached mid-to-late exponential phase. GFP fluorescence (left), mCherry fluorescence (right and middle) and merged (far right) images are overlaid with phase-contrast images of the same fields. Inserts are magnified views of single cells, showing that GFP fusion foci did not overlap with inclusion bodies even when both were present at same pole. B. Cells expressing GapA–GFP served as a non-polar control. GFP fluorescence and phase-contrast images were overlaid. C. IbpA–mCherry detects inclusion body formation. E. coli GL60 cells expressing IbpA–mCherry were incubated in the presence or absence of kanamycin. Fluorescent foci mark the sites of misfolded protein aggregates and inclusion bodies (Lindner et al., 2008). mCherry fluorescence and phase-contrast images were overlaid.

Overexpressed or misfolded fusion proteins may form inclusion bodies that are often deposited at the cell poles by nucleoid occlusion, by a poorly understood energy-dependent process, or due to cell division (Lindner et al., 2008; Rokney et al., 2009; Winkler et al., 2010). To rule out the possibility that polar foci were inclusion bodies, we employed the IbpA–mCherry protein which binds to misfolded proteins and labels inclusion bodies in vivo, including aggregates of fluorescently tagged fusion proteins (Lindner et al., 2008; Michaelis and Gitai, 2010; Van der Henst et al., 2010). Under non-stressed growth conditions (e.g. at 30°C or 37°C in 1% tryptone medium without antibiotics), cells containing IbpA–mCherry exhibited no detectable fluorescence (Fig. 3C, without kanamycin). However, even in resistant strains the addition of kanamycin causes problematic protein translation and provokes the production of IbpA–mCherry (Lindner et al., 2008). In cells stressed in this manner, IbpA–mCherry formed scattered midsize foci or large polar aggregates, marking the location of inclusion bodies composed of misfolded proteins (Fig. 3C, with kanamycin). This confirmed the previously reported behaviour of IbpA.

We next determined if IbpA–mCherry colocalized with the polar foci of Aer–GFP, YqjD–GFP, TnaA–GFP or GroES–GFP, which would indicate that these foci were inclusion bodies. When cells containing each candidate protein plus ibpA–mCherry were grown in non-stressful conditions, no IbpA–mCherry signal was detected in spite of the fact that the candidate fusion proteins formed bright polar GFP foci (Fig. 3A, without kanamycin). This indicated that the proteins did not form aggregates of misfolded proteins or inclusion bodies. However, it was possible that these particular GFP fusion proteins might not induce efficient expression of ibpA–mCherry, which would produce an artificially negative result. To rule this out, we stressed cells containing the GFP fusion proteins by adding kanamycin to induce the production of IbpA–mCherry. Even when visible amounts of IbpA–mCherry were produced by this method, IbpA did not colocalize with the candidate proteins (Fig. 3A, with kanamycin). Surprisingly, even when foci of the candidate proteins and IbpA–mCherry did localize to the same pole, the IbpA–mCherry signal did not overlap that of the GFP tagged protein (Fig. 3A, inserts), as would be expected if the fusion proteins were misfolded inclusion bodies. The results strongly suggested that the polar foci of Aer–GFP, YqjD–GFP, TnaA–GFP and GroES–GFP were not inclusion bodies but, instead, reflected the natural locations of these pole-associated proteins.

One last possibility was that the localization of these newly found polar proteins depended on the presence of the Tar protein, which could imply that isolation of these proteins might be due to protein–protein interactions rather than to proximity capture within common membrane vesicles. However, when we grew E. coli in LB medium under conditions where the flagellar system (including the expression of Tar and Aer) was repressed, TnaA, GroES and YqjD exhibited the same localization patterns as those shown in Fig. 4 (not shown). These results indicate that the localization of latter three proteins did not depend on the presence of Tar.

Figure 4.

Polar foci of TnaA, GroES and YqjD at different points in the growth phase. Cells expressing TnaA–GFP (strain GL69), GroES–GFP (strain GL129) or YqjD–GFP (strain GL131) were grown in LB at 37°C as in Fig. 7. The percentage of the population that have polar or midcell foci (TnaA and GroES) or dominant polar foci (YqjD) are indicated below each panel. From 50 to 150 cells were counted for each sample.

TnaA, GroES and YqjD form polar foci by different means

It was interesting that TnaA–GFP and GroES–GFP formed only one polar focus per cell despite the fact that the two poles of E. coli are almost undistinguishable. To begin to understand the localization process, we observed the behaviour of these proteins during different growth phases (Fig. 4). In the case of TnaA–GFP, 35% of cells in early exponential phase had a single polar focus (Fig. 4, 1 h). In the mid- and late-exponential phases of growth, almost all cells contained a single TnaA–GFP focus, predominantly at one pole, although some foci appeared at midcell in dividing cells (Fig. 4, 2 h and 3 h). The fraction of the population with TnaA–GFP foci decreased significantly when the cells entered into stationary phase, dropping to 58% at 6 h and to 31% at 16 h, and yet those foci that were still visible continued to be located predominately at one pole (Fig. 4). Apparently, TnaA foci assembled predominately during the mid-exponential phase of growth but slowed or ceased altogether when cells entered into stationary phase. Interestingly, in 10–30% of dividing cells, one daughter contained a large TnaA–GFP focus while the other daughter contained a much smaller protein mass (Fig. 5). Generally, the small TnaA–GFP foci formed at three locations: at the old pole, at the new pole or at midcell (Fig. 5). Approximately 85% of the small foci were observed at one or the other pole, with the remaining 15% at midcell. The location of small foci showed no apparent preference for either the old or the new pole. We note that it was not possible to distinguish clearly the origins of these foci; some nascent polar foci might have originated at midcell in a previous generation but localized to new poles after division.

Figure 5.

Origin of TnaA and GroES polar foci. Percentages of nascent (smaller) TnaA foci observed at the old pole, at the new pole or at midcell. Note that nascent polar foci might have originated at midcell in a previous generation. Nascent GroES foci were located most often at midcell.

Regarding the localization of GroES–GFP, all cells contained a single focus throughout growth, with very few exceptions (Fig. 4). Generally, large foci were present at one pole and small foci were present at midcell (Fig. 5), suggesting that foci formed at midcell and either stayed at the new pole or moved to the old pole after cell division. In contrast, almost all cells had one large YqjD–GFP cluster at a single pole when in stationary phase (Fig. 4, 16 h), but the numbers of polar foci plummeted to 85% and 15% as the cells entered early and mid-log growth respectively (Fig. 4, 1 h and 2 h). YqjD–GFP foci reappeared as scattered clusters during mid-to-late exponential phase (Fig. 4, 3 h and 6 h). Thus, the kinetics of polar localization differed among these three proteins.

Because TnaA and GroES formed polar foci mainly during logarithmic growth phase, we wished to know whether these foci contained a mixture of both proteins or whether the two were segregated into independent foci. To clarify this, we fused an mCherry-coding sequence to the chromosomal copy of groES to monitor the location of GroES–mCherry and TnaA–GFP in the same cells. The two fusion proteins localized to the same poles or to opposite poles, but even in cells where the proteins were at the same pole the two foci did not overlap (Fig. 6A). These results indicated that GroES–mCherry and TnaA–GFP assembled as independent foci that did not mix with one another. In addition, deleting tnaA had no effect on the distribution of GroES–GFP polar foci (Fig. 6B), indicating that localization of GroES did not depend on the presence of TnaA.

Figure 6.

Independent localization of TnaA and GroES. A. TnaA and GroES form independent unipolar foci. E. coli GL135 cells coexpressing chromosomal TnaA–GFP and GroES–mCherry were grown in LB to mid-to-late exponential phase. TnaA–GFP foci (green fluorescence in the left panel); GroES–mCherry foci (red fluorescence in the middle panel); and merged images of both types of foci (right panel). Phase-contrast images were overlaid with the respective fluorescence images. Foci of the two proteins did not overlap. Note that the two types of foci are separate entities so that one protein may be pushed to the side if the pole is occupied by a large focus of the other protein. B. GroES–GFP expression in a strain (GL346) from which tnaA was deleted. The fluorescent GFP image is overlaid with the phase-contrast image. C. Wild-type TnaA and GroES compete for polar localization with their respective GFP fusion proteins but not with one another. Strains expressing TnaA–GFP or GroES–GFP from the chromosome (strains GL69 and GL129 respectively) were transformed with plasmids carrying untagged wild-type TnaA and GroES proteins (pTnaA and pGroES respectively) or the vector control plasmid (Vector). Expression of the wild-type proteins was induced by adding IPTG (200 µM IPTG) and the cultures were grown in LB medium until they reached mid-to-late exponential phase. D. Longer induction times are required for GroES–GFP to form foci in the presence of wild-type GroES. Plasmid-encoded GroES–GFP was expressed in E. coli MG1655 by adding either 10 or 100 µM IPTG for 3, 4 or 5 h. The host cell expressed normal amounts of unlabelled GroES from the chromosomal locus. The percentage of polar or midcell foci in each sample was calculated by examining 50–100 cells.

TnaA–GFP and GroES–GFP each exhibited unipolar localization, so they could either recognize a common polar signal or be drawn to the poles by different mechanisms. We reasoned that if the two proteins recognized different but specific signals then untagged wild-type proteins should compete with their homologous GFP fusions for polar assembly sites but should not compete with the unrelated protein. We therefore expressed plasmid-borne, untagged TnaA or GroES in strains carrying chromosomally encoded TnaA–GFP or GroES–GFP. When untagged TnaA was overexpressed, the TnaA–GFP signal became completely diffuse in 97% of the population (Fig. 6C). When untagged GroES was overexpressed, the GroES–GFP signal was completely diffuse in 40% of the population, with the remainder of cells exhibiting lighter foci and a much increased background (Fig. 6C). In contrast, expression of untagged TnaA did not affect the polar localization of GroES–GFP, and untagged GroES did not affect the polar localization of TnaA–GFP (Fig. 6C). The results indicated that the localizations of TnaA and GroES were independent and unaffected by the presence of one another, even when expressed in excess.

The unipolar localization we observed for GroES was at odds with results of previous localization experiments for GroES and GroEL (Ogino et al., 2004; Carrió and Villaverde, 2005; Cava et al., 2008; Winkler et al., 2010; Charbon et al., 2011). In particular, recent reports indicated that both GroES–GFP and GroEL were spread diffusely and evenly in the cytoplasm (Cava et al., 2008; Charbon et al., 2011). However, in these studies, tagged GroES and GroEL were expressed from plasmids in strains that retained the wild-type groES and groEL genes in the chromosome. To address the discrepancy between our results and previous work, we examined the localization of cloned GroES–GFP in the presence of the wild-type protein expressed from the chromosome. We induced GroES–GFP expression with 10–100 µM IPTG for 3 h, thereby producing protein under conditions similar to those of Charbon et al. (2011). Here, GroES–GFP did not localize uniquely to the poles but was, in the main, diffuse in the cytoplasm, as reported by Charbon et al. (Fig. 6D, left two panels). However, when the protein was expressed for 4 or 5 h, more than 50% of cells had polar and midcell GroES–GFP foci (Fig. 6D, right two panels), confirming that polar localization of GroES–GFP could still occur. The fact that a large amount of fusion protein was required to see polar localization in the presence of wild-type protein suggests that untagged GroES could have skewed the localization results of previous reports. Such a prospect is also supported by the GroES versus GroES–GFP competition results described above (Fig. 6B).

TnaA and GroES GFP fusions are functional in vivo

To evaluate the significance of the localization results, it was important to establish that each fusion protein was functional in vivo. We therefore tested the in vivo activity of TnaA and GroES, but did not test YqjD–GFP because its function is unknown. TnaA is a tryptophanase that cleaves the indole moiety from tryptophan (Watanabe and Snell, 1972), so we tested TnaA–GFP activity by measuring the concentration of indole in the culture medium during growth (Domka et al., 2006). All tested strains grew at the same rate (Fig. 7A), so differences in indole accumulation were not due to growth defects. Beginning at mid-log phase (3 h growth) the parent strain expressing wild-type TnaA accumulated a high level of indole (Fig. 7B, white bars), but a strain from which the tnaA gene had been deleted produced very little or none (Fig. 7B, black bars). This indicated that the appearance of indole depended on the presence of a functional TnaA protein. A strain lacking wild-type TnaA but expressing TnaA–GFP produced large amounts of indole that closely followed the kinetics of the parent strain (Fig. 7B, grey bars), indicating that the fusion protein was just as active as the wild-type protein. It was possible that the observed tryptophanase activity of TnaA–GFP might be due to proteolytic removal of the GFP domain, thereby releasing active wild-type TnaA. However, when we analysed cells by SDS-PAGE it was clear that the fusion protein was stable and that the GFP domain was not cleaved off in appreciable amounts, even after overnight culture (Fig. 7C). This indicated that wild-type TnaA did not account for the high level of indole production during growth and established that TnaA–GFP retained its enzymatic activity.

Figure 7.

TnaA–GFP and GroES–GFP are stable and functional. A. Wild-type and mutant strains grow at the same rate. MG1655, wild-type cells (open bars); TnaA–GFP, a strain lacking wild-type tnaA but expressing TnaA–GFP from the original chromosomal locus (grey bars); ΔtnaA, a strain (MGΔtnaA) from which the tnaA gene was deleted (black bars). Cells were grown in LB medium at 37°C. B. Indole concentrations of cultures shown in (A). The culture medium was assayed for the presence of indole to confirm TnaA–GFP tryptophanase activity. The notations are the same as in (A). Values are the averages from three independent cultures, with standard deviation as indicated. C. In-gel GFP fluorescence of cells expressing TnaA–GFP (from strain GL69) and GroES–GFP (from strain GL129) for 1–16 h (lane labels). After SDS-PAGE, the gel was washed and incubated in PBS to allow the GFP to refold. E. coli MG1655 expressing GFP from the plasmid pLP9 was loaded in the far right lane as a control. Bands representing the GFP fusion proteins are indicated by arrows.

As for the activity of GroES, it is part of the essential GroES–GroEL chaperonin that is involved in folding hundreds of proteins and is required for growth at all temperatures (Fayet et al., 1989; Xu et al., 1997; Kerner et al., 2005; Chapman et al., 2006). We found that the GroES–GFP fusion protein must have retained normal function because no growth defect was observed when the chromosomal groES gene was replaced by the groES–gfp construct (not shown). Thus, the GFP domain did not affect the activity of GroES or its ability to form a complex with GroEL. GroES–GFP was also stable during cell growth (Fig. 7C), arguing further that GroES–GFP was functional in vivo.

MinCDE proteins influence polar localization of TnaA, GroES and YqjD

Because some TnaA–GFP and GroES–GFP foci were observed at midcell, we thought that cell division might play a role in localizing these proteins. The MinCDE proteins oscillate from pole-to-pole to ensure that the FtsZ pre-division ring assembles at midcell (Margolin, 2005; de Boer, 2010). To see if interfering with FtsZ ring placement affected the localization of TnaA–GFP, GroES–GFP or YqjD–GFP, we observed the distribution of these proteins in a strain from which minCDE had been deleted. As expected, the cells divided not only at their midpoint but also at poles, this latter event producing small spherical minicells and long rods. In the absence of the Min system, TnaA–GFP, GroES–GFP and YqjD–GFP still formed foci. However, these foci were distributed irregularly along the lengths of each cell and did not localize expressly to cell poles or to nascent septa (Fig. 8). These results suggested that the Min system helped localize these proteins. The exact mechanism is under investigation.

Figure 8.

Loss of the MinCDE proteins disrupts polar localization of TnaA, GroES and YqjD. TnaA–GFP was expressed from the chromosome in E. coli MG1655 (top row, strain GL69) or in a strain from which the minCDE genes were deleted (second row, strain GL75). GroES–GFP and YqjD–GFP were also expressed in the ΔminCDE strain (third and fourth rows, strains GL147 and GL 149 respectively). Cells were grown in LB at 37°C until the cultures reached mid-to-late exponential phase, at which time membranes were stained with FM4-64 (red, all columns) and nucleoids were stained with DAPI (blue, centre column). GFP fluorescence images (left column), DAPI images (middle column) and the merged images (rightmost column) are overlaid with the FM4-64 images. Note the small spherical minicells at the poles of some rod cells. Foci of each of the three protein fusion proteins were distributed irregularly in the ΔminCDE cells, with a possible preference for localizing in nucleoid-free regions.

In addition to MinCDE, the SlmA nucleoid exclusion protein also restricts the positioning of the FtsZ ring (Bernhardt and de Boer, 2005). However, the polar localizations of TnaA–GFP, GroES–GFP and YqjD–GFP were not affected in strains from which slmA was deleted (not shown). Thus, either SlmA has no effect on the localization of these proteins or else the Min proteins play a more important role than does SlmA.


Despite the rapidly growing number of components found at bacterial poles, our knowledge about this area of the cell remains seriously incomplete. A prominent shortcoming is that there is no comprehensive inventory of proteins that localize uniquely or preferentially to the poles. Also, we have little information regarding the mechanisms responsible for drawing proteins to the pole, the means by which they are retained once there, or whether polar localization might be a transient or intermediate stop for some molecules. Finally, for a great many proteins, we do not know the biological or biochemical reasoning that would explain why they should be localized to the poles instead of elsewhere. Here, we describe a strategy for compiling a more wide-ranging list of bacterial pole-specific proteins, whether derived from the inner membrane, periplasm or cytoplasm, with the goal of generating a more systematic understanding of cell pole biology.

We enriched pole-specific proteins by affinity capture of membrane vesicles containing the Tar protein, with the idea that these vesicles would also contain wild-type versions of polar components in the vicinity of this chemotaxis component. The success and specificity of the technique was corroborated in several ways. First, the total of 124 proteins present in vesicles (polar plus non-polar) compares favourably to the 115 proteins observed in a two-dimensional PAGE/MS catalogue of E. coli inner membrane complexes (Stenberg et al., 2005).

In addition, our procedure identified 30 (71%) of the 42 proteins that were also identified by mass spectrometry by Stenberg et al. (Table S2). These correlations suggest that the affinity-capture procedure was able to isolate and identify the most prominent inner membrane proteins. Second, of 31 proteins that were highly enriched in polar vesicles, five were chemotaxis proteins already known to localize to the pole. These served as positive internal controls, confirming and highlighting the specificity of the affinity technique in being able to distinguish pole-derived membrane vesicles from those originating from other parts of the cell. In addition, the polar vesicles contained two proteins (PBP1B and Pal) that associate with the invaginating septum and which probably remain at the new poles after division is complete (Bertsche et al., 2006; Gerding et al., 2007). These latter proteins acted as an additional pair of internal controls, and their appearance in polar vesicles adds credence to previous speculations about their localization. Finally, as proof of concept, we characterized four new candidate polar proteins and confirmed that tagged versions of Aer, YqjD, TnaA and GroES did in fact localize to the poles in vivo, as predicted. Overall, the results strongly validate the efficacy and usefulness of the vesicle-capture strategy for isolating polar membrane vesicles and identifying new polar proteins.

Advantages and limitations of the affinity-capture strategy

The vesicle-capture strategy complements previous approaches for finding polar proteins and, in addition, has several advantages. The first advantage is that by analysing whole vesicles the approach expands the types of proteins that can be screened to include those membrane-associated proteins that reside in the cytoplasm and periplasm. Second, the strategy captures wild-type proteins that localize to the pole. The polar marker protein used for the vesicle capture stage can be any protein, natural or tagged, but in the vicinity of the marker the neighbouring proteins are unmarked and expressed in wild-type amounts, thus avoiding possible artefacts of protein engineering and overexpression. Third, proteins that are not permanent components of the pole but which localize to that region for even brief periods of time may be identified by enrichment with respect to their concentrations in other parts of the cell envelope. For example, MinD is a membrane-binding protein that oscillates from pole to pole but is associated transiently with sidewall membrane along the cell cylinder and near the poles. The polar score for MinD was ‘3’, which reflects fairly well its principle relationship to the poles versus sidewall. So it seems the affinity-capture method has the level of sensitivity required to identify a proteins that localize to the poles only part of the time. Finally, those proteins that localize temporarily at septation sites during cell division have time to diffuse away, thereby reducing contamination by proteins that are not bona fide components of the poles. An additional advantage is that the approach can identify proteins that avoid the pole, and the mechanisms that exclude these proteins should be equally informative for understanding how bacterial cells are organized.

One major limitation affects all MS-based procedures in that inner membrane proteins are often refractory to MS analysis by virtue of having a limited number of proteolysis sites (Weiner and Li, 2008). The magnitude of this limitation may be large because the inner and outer membranes of E. coli contain an estimated 1000 (Diaz-Mejia et al., 2009) or 1200–1700 proteins (Poetsch and Wolters, 2008). The fact that we identified only 124 proteins in our samples suggests that many may remain to be identified and their localizations explored. However, we note that these large numbers are only estimates derived from bioinformatic analyses, which do not take into account the fact that not all proteins are expressed constitutively. In addition, a limitation of the specific approach we used here is that Tar is an inner membrane protein, which eliminates the likelihood of finding pole-specific outer membrane proteins. However, the procedure can be modified to use an outer membrane protein or other known polar component as the affinity target. This would expand the candidate list to include those from other cellular compartments. In short, the affinity-capture procedure complements previous approaches for finding pole-specific proteins, avoids several limitations of earlier screens, and can be extended by choosing different affinity targets.

Comparisons with other polar localization approaches

Previous screens for finding pole-specific proteins have their own advantages and drawbacks. For example, the minicell-enrichment approach of Lai et al. is based on the sound reasoning that polar proteins will probably accumulate to a higher degree in these particles (Lai et al., 2004). As practiced, the procedure enriches certain types of pole-specific proteins, particularly those in the outer membrane. However, the approach as described has some limitations that reduce its ability to identify a broader range of pole-specific proteins. For example, only integral membrane proteins were screened because peripheral proteins were removed with a carbonate wash, thereby excluding pole-specific proteins that may reside predominantly in the cytoplasm or periplasm (Lai et al., 2004). Also, cytoplasmic membrane proteins were not well represented in the candidate list because these proteins are, in general, not separated efficiently by isoelectric focusing (Weiner and Li, 2008). Lai et al. pointed this out by noting that their 2D-gel procedure failed to detect several known pole-specific inner membrane chemotaxis proteins, including Tar, although these could be detected by Western blotting in one dimensional gels (Lai et al., 2004). In addition, not all proteins enriched in minicells need be pole-specific. For example, some cell division proteins may become trapped in minicells when septation is completed but such proteins would not necessarily be pole-specific. Finally, because these cells lack MinCDE, any protein whose polar localization depends on a functional Min system will be missed by this approach. This is relevant because we observed that the Min system helps localize at least three of our newly identified polar proteins.

Even though our vesicle-capture strategy and the minicell-enrichment approach of Lai et al. employ different ways of enriching polar components, the two techniques share some important findings that reinforce the usefulness of each (see Table S3 for a detailed comparison). First of all, both procedures detected chemoreceptors and Pal in pole-enriched fractions. Lai et al. also showed that the hypothetical protein YiaF was distributed in a patchy way throughout each cell but still exhibited a polar bias. Similarly, we found that YiaF was enriched threefold in polar vesicles (Table S1). This level of enrichment was below our (arbitrary) cut-off value for looking in more detail at candidate proteins, but the fact that both procedures detected a polar bias suggests that the pattern may be real and worth looking at. This type of result also implies that YiaF and perhaps other proteins may have something other than an ‘all or nothing’ affinity for the pole; instead, these may exist in a more interesting equilibrium with different parts of the cell. We did not identify OmpW as a polar protein as was observed by Lai et al., but the results can be reconciled because OmpW is an outer membrane protein and our affinity-capture approach targeted primarily inner membrane vesicles. Similarly, the four polar proteins we found were not identified by the minicell-enrichment procedure. However, these are inner membrane or cytoplasmic proteins that were absent or poorly represented in the minicell preparations (Lai et al., 2004). Thus, overall, the results reinforce the utility and complementarity of the two procedures.

On the other hand, some discrepancies exist between the results of our method versus the minicell approach. Lai et al. found that the ATP synthase components were enriched in minicells, suggesting that they may be polar. In contrast, we find these proteins to be evenly distributed, indicating a non-polar localization (Table 1). Similarly, the results for TolC were also different: although the protein did not accumulate in minicells, we observed a small TolC signal in polar vesicles (Table 2). That the TolC outer membrane protein was captured in our procedure, which targeted the inner membrane, is probably because TolC is part of several efflux pumps anchored to the inner membrane (Symmons et al., 2009). Why TolC was not enriched in the minicell procedure is not clear.

Strategies that rely on visual screens to find pole-localizing proteins have their own inherent drawbacks (Kitagawa et al., 2005; Janakiraman et al., 2009; Werner et al., 2009). For example, reliance on microscopy may fail to localize proteins that exist in very low concentrations. On the other hand, if tagged proteins are overproduced then false-positive polar localization may occur due to the presence of unfolded molecules or inclusion bodies. Using GFP as a tag for general screening creates other potential problems: e.g. the resulting fusion proteins may have low fluorescence or may inhibit cell growth if the domain interferes with protein folding or membrane translocation (Dinh and Bernhardt, 2011). Finally, in most of these screens tagged proteins are expressed in the presence of their untagged wild-type versions, and competition between the two forms may yield false-negative results (as we found to be true for TnaA and GroES). Thus, in many cases, the affinity-enrichment technique may be more sensitive for finding new types of polar proteins.

Reconciling conflicting observations regarding the cellular location of GroES and GroEL

We observed that GroEL was associated only with pole-derived membrane vesicles, although the raw signal was relatively low, but that GroES–GFP foci were localized strongly to the poles. These results conflict with those of other investigators who report that these proteins have a diffuse localization. In particular, GroEL has been shown to be distributed evenly throughout the cytoplasm in E. coli, whether assayed by electron microscopy (Carrió and Villaverde, 2005), immunofluorescence detection of the wild-type protein (Ogino et al., 2004; Winkler et al., 2010) or localization of an intrinsically fluorescent GroEL protein (Charbon et al., 2011). On the other hand, to our knowledge, only one study has addressed the localization of GroES, which was shown to be distributed evenly in the cytoplasm of Thermus thermophilus at 70°C (Cava et al., 2008). Although the GroES conflict may be explained by differences in the two organisms being observed (E. coli versus T. thermophilus), our GroEL results are more difficult to reconcile with previous observations.

A possible technical explanation for the localization discrepancies lies in the different protein expression strategies used among these experiments. We expressed GroES–GFP from its gene at the normal chromosomal site, under control of its native promoter, and in the absence of wild-type GroES. In contrast, previous studies expressed GFP-tagged GroES (Cava et al., 2008) and fluorescent GroEL (Charbon et al., 2011) from cloned genes in situations that required the fusion proteins to compete with untagged wild-type proteins. We showed that overproducing wild-type GroES removed GroES–GFP from the poles and, furthermore, that only by extended induction of cloned GroES–GFP could the tagged protein form polar foci in the presence of endogenous wild-type proteins. Thus, it seems likely that previous experiments may have created a competitive situation that eliminated or obscured polar localization, at least for GroES–GFP.

Because we tracked the localization of a GroES fusion protein, it could be argued that the GFP moiety affected the localization dynamics of GroES. However, this seems unlikely for several reasons: GroES–GFP was functional (cells grew normally in its presence, meaning it must interact with GroEL), the formation of polar foci was specific and did not interfere with polar foci composed of other proteins (e.g. TnaA), and wild-type untagged GroES competed with GroES–GFP for polar localization and vice-versa. Thus, the evidence suggests that GroES–GFP retained its function and localization.

A final possibility is that the apparently disparate results are not in conflict but reflect two true, but subtly different, localization behaviours of GroES and GroEL. The GroES–GroEL chaperonin is comprised of two heptameric GroEL rings stacked back-to-back, with a heptameric GroES cap sitting atop one of these rings (Hartl and Hayer-Hartl, 2002). However, GroEL can operate without GroES in some instances. Also, since GroES cycles on and off GroEL (Hartl and Hayer-Hartl, 2002), if GroES forms a polar focus then GroEL might be localized diffusely but interact intermittently with GroES at the poles. The fact that we observed a slight GroEL signal in polar vesicles may suggest the existence of two pools of GroEL, one polar and one diffuse, which would reconcile our findings with those of Winkler et al. and Charbon et al. (Winkler et al., 2010; Charbon et al., 2011).

Experimental procedures

Strains, plasmids and growth conditions

Strains and plasmids are listed in Table 3. Strains WM1032 and JW3686 were the sources for the ΔminCDE::kan and ΔtnaA::kan alleles respectively (Sun and Margolin, 2001; Baba et al., 2006). The vector plasmid pLP8 (KanR) carries the lac promoter and the lacIq gene (Potluri et al., 2010). Cells were grown in LB broth or in 1% Tryptone motility medium (tryptone, 10 g l−1; NaCl, 2.5 g l−1). Plasmids and strains were selected and maintained by adding kanamycin (Kan) (50 µg ml−1) when necessary. Expression of plasmid-encoded Tar-FLAG genes was induced by adding IPTG (25 µM final). Expression of plasmid-encoded TnaA, GroES or GroES–GFP was induced by adding 10–200 µM IPTG.

Table 3.  E. coli strains and plasmids.
Strain/plasmidDescriptionaSource or reference
  • a. 

    frt’ indicates the presence of the following oligonucleotide scar left after removal of the kan cassette: GAAGTTCCTATTCTCTAGAAAGTATAGGAACTTC

EC100F-mcrAΔ(mrr-hsdRMS-mcrBC) Φ80dlacZΔM15ΔlacX74 recA1 endA1 araD139Δ(ara, leu)7697 galU galK λ-rpsL (StrR) nupG (for plasmid construction)Epicentre Biotechnologies
MG1655Wild-type F- lambda-ilvG-rfb-50 rph-1Laboratory collection
JW3686BW25113 ΔtnaA::kanKeio collection
MGΔtnaAMG1655ΔtnaA::frtD. Vega (unpublished)
WM1032MG1655 ΔminCDE::kan Sun and Margolin (2001)
GL60MG1655 ibpA–mCherry::kanThis study
GL69MG1655 tnaA–GFP::frtThis study
GL75GL69 tnaA–GFP::frtΔminCDE::kanThis study
GL81GL69 tnaA–GFP::frt ibpA–mCherry::kanThis study
GL129MG1655 groES–GFP::frtThis study
GL131MG1655 yqjD–GFP::frtThis study
GL135GL69 tnaA–GFP::frt groES–mCherry::kanThis study
GL137MG1655 groES–mCherry::kanThis study
GL143GL129 groES–GFP::frt ibpA–mCherry::kanThis study
GL145GL131 yqjD–GFP::frt ibpA–mCherry::kanThis study
GL147GL129 groES–GFP::frtΔminCDE::kanThis study
GL149GL131 yqjD–GFP::frtΔminCDE::kanThis study
GL197MG1655 aer–GFP::frtThis study
GL199GL197 aer–GFP::frt ibpA–mCherry::kanThis study
GL346GL129 groES–GFP::frtΔtnaA::kanThis study
pLP8PlaclacIq KanR Potluri et al. (2010)
pLP9 dsbA(SS)-sfGFP in EcoRI/HindIII of pLP8 Potluri et al. (2010)
pLP14 mCherry in BamHI/HindIII of pLP8L. Potluri (unpublished)
pDoc-sfGFP sfGFP::kan in pDOC-KL. Potluri (unpublished)
pTar-P-FLAG tar-P-FLAG in EcoRI/BamHI of pLP8This study
pTar-C-FLAG tar-C-FLAG in EcoRI/BamHI of pLP8This study
pTar-P-FLAGΔ tar-P-FLAGΔ258–553–GFP in EcoRI/BamHI of pLP8This study
pTnaA tnaA in EcoRI/BamHI of pLP8This study
pGroES groES in EcoRI/BamHI of pLP8This study
pGroES–GFP groES–GFP in EcoRI/BamHI of pLP8This study

DNA purification and manipulation

DNA fragments and plasmids were purified by using QIAquick PCR or gel extraction kits and the QIAprep miniprep kit (Qiagen). Plasmids were transformed by electroporation into E. coli EC100 or MG1655. All plasmids were sequenced to avoid unwanted mutations.

Plasmid constructions

The DNA fragment encoding Tar-P-FLAG was generated by sequential PCR, as described (Cormack, 2001), by using chromosomal DNA from MG1655 as a template. The tar-C-FLAG fragment was generated by PCR, with the FLAG sequence included in the reverse primer. The tar-P-FLAGΔ fragment was generated by sequential PCR, with pTar-P-FLAG as one template and pLP9 as the GFP fragment template. All GFP proteins were the superfolder GFP (sfGFP) version obtained from pLP9 (Pedelacq et al., 2006; Potluri et al., 2010). The tnaA and groES DNA fragments were amplified from the chromosome of E. coli MG1655. The groES–GFP DNA fragment was amplified from the chromosome of GL129. PCR fragments were digested with EcoRI and BamHI, and ligated to the corresponding sites in pLP8. All primers are listed in Table S4.

Construction of chromosomal GFP and mCherry gene fusions

Chromosomal C-terminal GFP and mCherry fusion proteins were constructed by inserting a GFP::kan or an mCherry::kan cassette immediately prior to the stop codons of each target gene, by using λ-Red recombination (Datsenko and Wanner, 2000). All primers are listed in Table S4. Primers used to amplify the GFP::kan or mCherry::kan cassette had 40 nt sequences that were homologous to the C-terminal regions and the downstream regions of the target genes. To construct aer–GFP, the GFP::kan cassette was amplified from pDoc-sfGFP. pDoc-sfGFP was constructed by inserting the sfGFP gene fragment into the XmaI and KpnI sites of pDoc-K (Lee et al., 2009). To construct other GFP fusion proteins, GFP::kan cassettes were amplified from aer–GFP::kan. To construct ibpA–mCherry, the mCherry::kan cassette was constructed by sequential PCR using pLP14 as the template for the mCherry fragment and by using pDoc-sfGFP as the template for the kan fragment. To construct groES–mCherry, the mCherry::kan cassette was amplified from ibpA–mCherry::kan. Chromosomal insertions were confirmed by diagnostic PCR, after which the kan cassette was removed by using FLP recombinase as described (Datsenko and Wanner, 2000).

Immunolocalization of Tar-P-FLAG

Overnight E. coli MG1655 cells containing pTar-P-FLAG were diluted 1:100 into 1% Tryptone medium and incubated at 30°C. After 1 h, expression of Tar-P-FLAG was induced by adding 25 µM IPTG and incubation was continued for an additional 3 h at 30°C. Subsequent steps were performed at room temperature. Cells were fixed for 15 min with 2.8% formaldehyde and 0.04% glutaraldehyde, pelleted, washed twice with PBS, and permeabilized by suspending for 15 min in 50 mM Tris pH 8.0, 5 mM EDTA and 100 µg ml−1 lysozyme. Cells were washed twice in PBS plus 0.05% Tween 20 (PBST), resuspended in PBST plus 1% BSA and incubated for 30 min. Anti-FLAG M2 conjugated with FITC (Sigma-Aldrich) was added to give a final concentration of 5 µg ml−1. After 1 h incubation, cells were washed twice with PBST and loaded onto agarose-coated slides for microscopy.

Isolation of inner membrane vesicles

Escherichia coli MG1655 containing one of the Tar-FLAG plasmids was grown in 100 ml of 1% Tryptone medium and expression of one of the Tar-FLAG proteins was induced as described for the immunolocalization of Tar-P-FLAG (above). Cells were harvested, resuspended in 15 ml of PBS, and shaken at 30°C for 1 h to allow newly synthesized Tar proteins localize to the poles. After adding 60 µl of Halt protease inhibitor (Thermo Scientific), RNase (16 µg ml−1) and DNase (1.3 µg ml−1), cells were disrupted by passing the suspension through an LV1 high shear fluid processor (Microfluidics) at 20 000 psi. After a low-speed spin (10 000 g for 5 min) to remove unbroken cells and large envelope fragments, 10 ml of lysate containing the membrane vesicles was mixed with 2 ml of PBST and 150 µl of anti-FLAG gel beads (Sigma-Aldrich), followed by incubating on ice for 1 h with slow shaking. The beads were washed twice with 8 ml of cold PBST, moved to 1.5 ml tubes and washed twice more. Washed beads were resuspended and incubated with 1% Triton X-100 in PBS (85 µl) at room temperature for 10 min to solubilize inner membranes (Schnaitman, 1971). After centrifugation to pellet the beads, proteins released into the supernatant were moved to clean tubes. The remaining beads were washed once again with 1% Triton X-100 in PBS and then resuspended in hot 1% SDS (85 µl). The samples were centrifuged to pellet the beads, and eluted proteins were collected from the supernatant. Protein concentrations were determined by using the Micro BCA protein assay kit (Thermo Scientific). Identical amounts of protein, except for the vector control containing contaminating proteins, were separated on 12% Mini-protean TGX SDS-PAGE gels (Bio-Rad) and stained with Sypro Ruby (Bio-Rad).

Full-lane GeLC-MS/MS

Mass spectrometry was performed by the University of Arkansas for Medical Sciences Proteomics Core Facility. Each SDS-PAGE lane was cut into 3 mm slices and subjected to in-gel trypsin digestion. Protein-containing gel slices were destained in 50% methanol (Fisher) and 100 mM ammonium bicarbonate (Sigma-Aldrich), followed by reduction in 10 mM Tris[2-carboxyethyl]phosphine (Pierce) and alkylation in 50 mM iodoacetamide (Sigma-Aldrich). Gel slices were dehydrated in acetonitrile (Fisher), 100 ng of porcine trypsin (Promega) in 100 mM ammonium bicarbonate (Sigma-Aldrich) was added, and the samples were incubated at 37°C for 12–16 h, after which the peptide products were acidified in 0.1% formic acid (Fluka). Tryptic peptides were separated by reverse phase HPLC on a 10 cm C18 column by using a NanoLC 2D system (Eksigent) and ionized by electrospray upon elution, followed by MS/MS analysis using collision induced dissociation on an LTQ XL mass spectrometer (Thermo). Proteins were identified from MS/MS spectra by database searching using the Mascot search engine (Matrix Science). Mascot search results were compiled using Scaffold (Proteome Software).

In-gel fluorescence imaging

In-gel fluorescence images of TnaA–GFP and GroES–GFP were obtained as described (Drew et al., 2006), except that cell samples were heated at 90°C for 3 min to denature the proteins before being separated on 12% Mini-protean TGX SDS-PAGE gels (Bio-Rad). After electrophoresis, the gels were washed three times with water and once with PBS and incubated in PBS for 1 h to allow superfolder GFP to refold. Gels were scanned for fluorescence by using a Typhoon TRIO scanner (GE Healthcare) set at 488 nm excitation and 520 nm emission.


Cells were visualized by phase-contrast and fluorescence imaging by using an Olympus microscope BX60 with a 100× oil objective (1.3 NA PH3), a 1.4 MP MONO CCD camera XM10, and illuminated by the fluorescence source X-Cite 120. Except for immunolocalization of Tar-P-FLAG, live cells were applied to 1% agarose-coated slides for imaging. When required, cultures were incubated with 0.5 µg ml−1 DAPI (Invitrogen) or 20 µg ml−1 FM4-64 (Invitrogen) at room temperature for 15 min before microscopy. GFP was visualized by using the OSF-0008Z filter set (471 nm excitation/520 nm emission), mCherry and FM4-64 were visualized by using the OSF-0025Z filter set (562 nm excitation/624 nm emission), and DAPI was visualized with the OSF-0007Z filter set (387 nm excitation). Images were captured by using CellSens software with exposure times of ∼ 300 ms, except that IbpA–mCherry-stained cells were exposed for 1 s.

Indole assay

Indole was measured as described previously, with minor modifications (Mueller et al., 2009). Culture supernatants (50 µl) were mixed with Kovac's reagent (1 ml) (Sigma-Aldrich) and incubated at room temperature for 1 h, after which the OD571 was measured. A range of indole (Sigma-Aldrich) solutions were used as standards (0–1 mM).


We thank L. Prasad Potluri for constructing the pLP series of plasmids. We also thank Dr Ikuro Kawagishi (Hosei University) for providing the Tar–GFP strain HCB436/pDS1030 that we used for preliminary Tar localization experiments. Funding for this project was provided by the US Government and managed by the Army Research Office under Award No. W911NF-10-1-0058. Additional support was provided by the Arkansas Biosciences Institute, the major research component of the Arkansas Tobacco Settlement Proceeds Act of 2000.